Self-Replicating Machines are the only means through which to spread through the Solar System

Alex Ellery Canada Research Professor Department of Mechanical & Aerospace Engineering Carleton University Ottawa, CANADA Space Portal Commercial Space Lecture Series, 2 Jun 2021 Power of Self-Replication

§ Self-replication is the ONLY robust approach to space exploration § Self-replication offers exponential growth in productive capacity P= (1 + �)! where r>1 § This is a cellular approach in which each cell factory is identical § Consider launch of a single 10 tonne self-replicating factory to the Moon at a cost of $7.5 B Number of offspring Number of Population Specific per generation generations Cost ($/kg) 1 10 1024 2 7 2187 <$1000/kg 2 13 1,594,323 <$1.25/kg § Cumulative population is >1.5 x 106 within 13 generations § Initial capital cost of $7.5 B is amortised over an exponentially increasing productive capacity § For r=2 over 13 generations, specific cost to the Moon has dropped from $750,000/kg to <50 cent/kg of productive capacity § If each 10 tonne factory takes 6 months to build, we have >1.5 million factories in 6.5 years § If each of the106 factories produces 103 solar power satellites, we have109 SPS units § We could supply our global energy needs cleanly from space § SELF-REPLICATION EFFECTIVELY SIDESTEPS LAUNCH COSTS Universal Constructor ⊃ Self Replicator

§ John von Neumann – a sufficient (but not necessary) condition for self-replication is universal construction § Universal constructor is a kinematic machine that can manufacture any other machine including a copy of itself § UC is generalisation of the universal Turing machine § UC is idealised as a programmable “generalised” kinematic arm in a sea of parts (structured environment) § We adapt UC to unstructured environments through a suite of kinematic machines: (i) load- haul-dumpers and drills as mining robots; (ii) programmable pumps for driving unit chemical processors; (iii) mills, lathes and 3D printers for autonomous manufacture; (iv) manipulators for robotic assembly § Kinematic machines are specific kinematic configurations of electric motors § Sensorimotor control system: - motors - computational electronics - sensors Functionality Lunar Material Tensile structures Wrought – Aluminium Minimal Demandite Compressive structures Cast iron – Aluminium Elastic structures Steel/Al springs/flexures Silicone elastomers Thermal conductor straps Iron/Nickel//Aluminium Tungsten Thermal insulation Glass (silica fibre)

Ceramics such as SiO2 and Al2O3 Moon has a simple geology: High thermal tolerance Tungsten, Al2O3 Thermal sources Fresnel lenses/mirrors (optical structures) pyroxene – plagioclase (anorthite) Electrical heating (iron/nickel/tungsten) – olivine - ilmenite Electrical conduction Fernico (e.g. ) Nickel – Aluminium Electrical insulation Glass (fused silica)

Ceramics - SiO2, TiO2, Al2O3 10 basic materials required to Silicone plastic supply all full functionality to Silicon steel for motors Active electronics Kovar robotic spacecraft (satisficibility) (vacuum tubes) Nickel Tungsten Fused silica glass/ceramic Magnetic materials Co-ferrite - AlNiCo Silicon steel We require a minimal set of reagents Permalloy resources including NaCl imported Sensors and sensory transduction Resistance wire Quartz from Earth – salt contingency Selenium Optical structures Polished nickel/aluminium (mirrors) Fused silica glass (lenses/fibres, etc) Lubricants Silicone oils Water Adhesives Silicone elastomer/gels

Fuels Oxygen – Hydrogen Electromagnetic launchers + solar sails 4 Self-Replication requires Closure

§ UC must be supplied with the appropriate resources to self-replicate: matter+energy+information § The most important constraints are closure conditions: (a) Material and parts closure – each component has a portfolio of materials, processes and machines required for its manufacture – to close the matter loop by: (i) Restricting raw material inventory to minimise mining and chemical processing cost, e.g. demandite list, industrial ecology, Metalysis FFC process (ii) Restricting parts inventory to minimise manufacturing and assembly cost, e.g. 3D printing (b) Energy closure – energy generation must exceed energy cost (energy return on investment (EROI) – to close the energy loop by: (i) Maximise use of direct environmental energy, e.g. thermal energy for chemical processing (ii) Maximising electrical energy conversion efficiency, e.g. PETE >30% (iii) Minimise energy wasted in processing waste by eliminating waste, e.g. industrial ecology (iv) Restrict all physical processes to minimise energy consumption, e.g. Metalysis FFC process, 3D printing § (c) Information closure – information required for specification does not exceed capacity to store/process specification - to close the information loop by: (i) Maximise parts re-use (exaptation) to minimise information specification, e.g. electric motors co-opted for energy storage as flywheels and vacuum tubes co-opted for energy conversion Lunar Mining

§ Processing of mineral in-situ resources requires mining vehicles § Kapvik is a 32 kg microrover prototype for lunar mining (i) JCB prototype with regolith scoop (ii) autonomous navigation capability using UKF implemented on two FPGAs (iii) reconfigurable chassis, e.g. elastic loop mobility system for high traction on rugged terrain (iv) sensorised chassis for online regolith characterisation (v) hybrid 4 DOF camera mast/manipulator arm (vi) pan/tilt camera at elbow with full observability of scoop (vii) abseiling capability for crater surveying Lunar Volatiles Volatile species Concentration Mass/m3 regolith (g)

(ppm/μg/g) assuming 1660

kg/m3 density

o § Volatiles mining by heating regolith to 700 C H 46±16 76 releases 90% of volatiles esp from smaller 3He 0.0042±0.0034 0.007 ilmenite particles (extracted magnetically): 4He 14±11 23

96% H2, <4% He, CO, CO2, CH4, N2, NH3, H2S, C 124±45 206

SO2, Ar, etc N 81±37 135 § Carbon compounds ~80-120 ppm F 70±47 116 Cl 30±20 50

§ Fractional distillation for well-separated fractions (assuming STP):

He (4.2 K), H2 (20 K), N2 (77 K), CO (81 K), CH4 (109 K),

N2O (185 K), CO2 (194 K), H2S (213 K), NH3 (240 K), SO2 (263 K)

NO2 (294 K) and H2O (373 K) § VERY VALUABLE STUFF!!!! § We should husband this stuff as we extract water ice…… Plastics/Ceramics Manufacture

§ Ceramics – fused silica glass – may be adopted for electrical insulation of cables § Ceramics – porcelain from fired kaolinite – may be adopted for electrical insulation of junctions § Knob-and-tube technology for electrical distribution § We adopt silicone plastics with multiple advantages over hydrocarbons (i) Si backbone minimises C resource consumption (ii) UV radiation resistant (iii) High operational temperature tolerance (350oC c.f. 120oC) (iv) it is used only sparingly for flexible wire sheathing § From syngas to polydimethylsiloxane (PDMS) – simplest siloxane (Rochow process) § HCl reagent is recycled – Cl required from “salt” contingency against uncontrolled replication § Primary use is for silicone oils Iron Age Technology

§ Hydrogen reduction of ilmenite at ~1000oC creates oxygen, iron and rutile

FeTiO3 + H2 → Fe + TiO2 + H2O and 2H2O → 2H2 + O2 o § Fe separated from TiO2 by liquation at 1600 C § Wrought iron is tough & malleable for tensile structures § Cast iron (~2-4% C + 1-2% Si) is more brittle for compressive structures (e.g. Iron Bridge for 200+y) § Tool steel (<2% C + 9-18% W) for cutting tools (eg. milling head) § Silicon (electrical) steel/ferrite (up to 3% Si and 97% Fe) for electromagnets and motor cores § Kovar (53.5% Fe, 29% Ni, 17% Co, 0.3% Mn, 0.2% Si and <0.01% C) – fernico alloy with high temperature electrical/thermal conductors 5 § Permalloy (20% Fe + 80% Ni) provides magnetic shielding with μr~10 H/m (replace 5% Ni 6 with Mo gives supermalloy with μr~10 H/m) Tunicose Ores from Meteoritic Sources

§ We need to source Tungsten, Nickel and Cobalt for our range § Some 25% lunar impactor material survives impact at or near surface of crater (670 crater >10 km diameter) § Mascons indicate location of NiFe meteorite ores, eg. northern rim of SPA crater § NiFe meteorites dominated by kamacite/taenite (NiFe alloys) - typically contaminated with Co § Mond (carbonyl) process at 40-80oC reacts impure Ni with CO and S catalyst o which is reversed at 230 C/60 bar: Ni(CO)4 ↔ Ni + 4CO § Similar carbonyl processes for Co and Fe § S catalyst recovered at 750-1100oC from troilite (FeS) in meteoritic inclusions, lunar regolith (~1%), or lunar volatiles (SO2 and H2S gases) § AlNiCo permanent magnets § Meteoritic NiFe alloys enriched in W microparticle inclusions which can be crushed and separated (W has high density of 19.3 and high melting temperature of 3422oC) Preprocessing the Natural Way

o § Carbothermic reduction of anorthite (CaAl2Si2O8) at 1650 C

yields Al2O3: o 4CH4 ® 4C + 8H2 (T=1400 C) o CaAl2Si2O8 + 4C → CaO + Al2O3 + 2Si + 4CO (T=1650 C) Alumina is refractory material (used in crucible linings) with physical properties exceeded only by diamond Quicklime is used as coatings for tungsten cathodes in vacuum tubes

§ Artificial weathering of anorthite (CaAl2Si2O8) with hot HCl yields SiO2 (artificial weathering)

CaAl2Si2O8 + 8HCl + 2H2O → CaCl2 + 2AlCl3.6H2O + 2SiO2

2AlCl3.6H2O → Al2O3 + 6HCl +9H2O Silica is raw material for fused silica glass (eg. Fresnel lenses) and quartz manufacture

CaCl2 electrolyte for FFC process as a byproduct § Quartz (piezoelectric) does not occur naturally on the Moon but it may be grown from silica o Melt silica at 2000 C followed by seeding in Na2SiO3 (Na2CO3+SiO2 → Na2SiO3+CO2 at 1700oC) at 350oC and 150 bar § Alumina and silica may be reduced to aluminium and silicon metals directly Electrolytic Metalysis FFC Process = Universal Chemical Processor

§ Cathode is sintered solid metal oxide (applicable to all metal oxides) – solid state reaction

MOx + xCa ® M + xCaO ® M + xCa + ½xO2

CaO + 2NH4Cl ® CaCl2 + 2NH3 + H2O o § CaCl2 electrolyte at 900-1100 C with O2 evolved at the anode (assumed non-eroding) - graphite anode yields CO/CO2 – recyclable through Sabatier process) § Product is >99% metal alloy sponge that can be powdered for 3D printing

TiO2 powder → Ti powder → 3D printed Ti parts § Output powder has been directly 3D printed into Ti test parts with SLS § Metalysis FFC process requires 97% thermal heating and 3% electrolytic energy Lunar Industrial Ecology 3D Printing = Universal Construction Mechanism

§ RepRap FDM 3D printer can print many of its own plastic parts

§ Full self-replication requires 3D printing: (i) structural metal bars and components (SLS/M or EBAM) (ii) joinery (replaced with cement/adhesive) (iii) electric motor drives (iv) electronics boards (v) computer hardware/software

§ Full self-replication also requires: (i) self-assembly (proxy for manipulator motors) (ii) self-power (solar-thermionic/flywheel) (iii) material processing into feedstock - ISRU

From 3D printed electric motors and electronics, omnia sequitur…

Moses et al concept of universal constructor Fully 3D Printed Motor

§ 3D printed rotor (ProtoPasta) § 3D printed permanent stator magnet (Oak Ridge National Laboratory) § LOM-style copper tape wiring/commutator wound around rotor § 3D printed shaft + bearings Photo-Sensitive Materials

§ Imaging camera array is fundamental as a remote distance sensor § P-type semiconductor selenium was used in Victorian photophone § Se is found in association with metal sulphides but, although rare on the Moon, chalcopyrite (Cu-Fe-S) deposits exist § S/Se ratio in carbonaceous meteorites is ~2450 with S~5% content of same § In NiFe meteorites, Se is associated with troilite (FeS) and chalcopyrite (Cu-Fe-S) deposits

§ Iron selenide may be smelted with soda Na2CO3 and saltpetre KNO3:

FeSe + Na2CO3 + 1.5O2 ® FeO + Na2SeO3 + CO2

§ Selenite Na2SeO3 is acidified with H2SO4 that precipitates tellurite impurities out of solution leaving selenous acid (H2SeO3) from which Se may be liberated:

H2SeO3 + 2SO2 + H2O ® Se + 2H2SO4 § Se is recovered with sulphuric acid recycled

§ Imported reagent Na is necessary for Se extraction as part of the `salt` contingency (Na2O product is recycled by water into NaOH + HCl ® NaCl +H2O) § Photomultiplier is a with electrodes coated with Se (primary emitter) and Al2O3/MgO (secondary emitters) Generalised 3D Printing Suite

§ 3D printing offers versatility in manufacturing 3D structures All 3D printers = XY printing head on Z deposition table = cartesian robots § We are interested in two technologies – selective solar sintering and EBAM § Selective solar sintering uses solar concentrators for generating thermal energy § We have a 2m x 2m diameter Fresnel lens capable of generating up to ~1500oC spot temperature § Quartz rod → fibre-optic cabling → selective solar sinterer § EBAM is electron gun (high voltage vacuum tube) but restricted to metals § Wire-fed EBAM (NRC) can print Ti alloy striucture and Al alloy wiring Multi-Material 3D Printer

§ We are building rigid multi-material 3D printer to print in metals and plastics § Solar furnace based on 1.2 m x 0.9 m Fresnel lens for melting Al alloy in ceramic crucible → 3D printing by Fresnel lenses § We have deposited molten Al wire tracks onto silicone plastic insulation: Al alloy (440oC m. p.) ↔ silicone (350oC op temp) § We have demonstrated metal deposition on plastic!

§ Two heads being added – fibre optic and silicone plastic extrusion head § Phase 2 - 3rd milling head for integrated surface finishing § Phase 3 – 4th wrist assembly head for component assembly § Phase 4 – Migrate to steels/silicone-derived ceramics

SiOxCy + (1-x+2y)O2 → SiO2 + yCO2 § Initially, we shall 3D print passive electronic circuitry Steam-Punk Electronics

§ Vacuum tubes are still used for TWTAs on spacecraft (more rad-tolerant than solid state) § Vacuum tubes are of simple construction that are potentially printable (TBD) § Thermionic devices which use kovar resistance wire to heat a sintered tungsten cathode coated with CaO to 1000-2000oC which evaporates electrons to Ni anode and control grid in a evacuated glass envelope § Example: Colossus at Bletchley Park (1943) of 2400 vacuum tubes designed by the great and unsung Tommy Flowers § Problem: circuit complexity growth with task program, e.g. ENIAC with 17,000 vacuum tubes occupied a large room § CPU-based architectures grow exponentially, e.g. 8086 CPU architecture is an early simple von Neumann architecture § Modern CPUs are highly complex – cannot be implemented using vacuum tubes § Solution: (i) General purpose computer architecture can be replaced with distributed architecture of specialised circuits (ii) Artificial neural nets grow logarithmically with task size (Parberry) Analogue Neural Nets = Turing Complete

§ Recurrent neural nets are Turing-equivalent (indeed, analogue neural nets offer super-Turing capabilities) § Direct model of original Turing machine: Input tape = magnetic core memory (same components as motor) Output tape = analogue neural net circuits Read/write head = 3D printer § Modified Yamashida-Nakaruma hardware “printable” neuron with fixed weights § MLP neural net circuit with BP circuit for obstacle avoidance:

§ Weights can be updated but not topology – limits on learning § Implementation of RatSLAM in neural network based on the Rat § hippocampus? Magnetron

§ We intend to 3D print a magnetron – it is a macroscopic vacuum tube with “motor” elements + cooling fins § We wish to explore direct thermal heating of the cathode using solar concentrators to eliminate electrical conversion stage § Magnetron is the centrepiece of the SPS § Magnetron introduces further capabilities for self-replication: (i) regolith processing; (ii) pn junction doping; (iii) scientific analysis instruments (MIBS) Electric Power Conversion

§ Thermionic conversion in vacuum tube with simple construction § Work function = min energy required to liberate electrons φ~2-6 eV (dependent on material) § For most materials, φ>3 eV so T >1000oC required, e.g. 4.52 eV for refractory tungsten requires T>1400oC

§ Alkaline earth oxide mixture BaO-CaO-Al2O3 in 4:1:1 ratio yields φ=2.87 – only CaO and Al2O3 available in-situ from anorthite § Al-doped haematite photocathodic coating? § To reduce space charge effect: (i) K vapour at > 760oC (similar low ionization potential as Cs) (ii) minimize inter-electrode distance ~1-10 μm (iii) electrostatic field by control grid electrode to shape electron transmission – η~40% § Photon-enhanced thermionic emission (PETE) with η=30-50% § Undoped Si has ΔE=1.1 eV offers good response § PETE may be supplemented by second stage thermoelectric conversion at anode

§ Thermoelectric Mg2Si manufactured by heating SiO2 with Mg powder (from olivine) § Conservative estimate η=30% (subject to detailed analysis/experiment) Flywheel Energy Storage

§ Electromechanical batteries (flywheels) offer zero DoD and insensitivity to charge/discharge cycling § High energy density ~100 kJ/kg and good power density ~50 Wh/kg § Energy stored, where w~20,000-50,000 rpm § Tangential velocity is determined by wheel material § Specific energy ~30 Wh/kg for steel – 40 Wh/kg for titanium – 100 Wh/kg for glass § To minimize radial stresses, rim constructed from concentric hoops separated by elastic material (e.g. silicone elastomer) § Halbach motor configuration – permanent magnet array in rotor - stationary coils in stator – magnetic bearings for frictionless operation § Brittleness of magnetic material to hoop stresses suggests use of magnetic composites comprising magnetic powder in a plastic matrix § Lunar-sourced magnetic material – AlNiCo/Co-ferrite permanent magnets, silicon steel/silica soft magnets, aluminium wiring and Ti/glass wheel structure Electromagnetic Launcher

§ Rather than burning lunar water (non-renewable resource) as propellant/oxidizer, electromagnetic launcher consumes only solar-derived energy § Electromagnetic launcher – coilgun is a rolled out linear DC motor

§ Carleton desktop e/m launcher built by Alex Craig-Sheldon Conclusions

§ I believe that we can make self-replication technology a technology of today not tomorrow § Am I mad as a hatter? § Answers to: Prof Alex Ellery Department of Mechanical & Aerospace Engineering Carleton University 1125 Colonel By Drive Ottawa ON. K1S 5B6 Canada § Tel: 1-613-261-3765 (M) or 1-613-520-2600 ext 1027 (W) § Email: [email protected] or [email protected]